A novel and useful method of using Incremental Connected Components to segment and isolate individual characters in a gray-scale or color image. For each pixel intensity of pixels in the image, a plurality of pixel groups are created comprising contiguous pixels of intensity equal to or less than the current pixel intensity. The pixel groups are then input to a character classifier which returns an identified character and a confidence value. Non-overlapping pixel groups (i.e. segmentation) of identified characters having the highest confidence values are then selected.
Legal claims defining the scope of protection, as filed with the USPTO.
1. A method of segmenting and isolating one or more characters from a non-binary image comprising one or more characters, said method comprising the steps of: identifying a pixel intensity value of each pixel in said non-binary image; generating a plurality of pixel groups comprising contiguous pixels of equivalent or lower pixel intensity values for each pixel intensity value found in said non-binary image, each such pixel intensity value defining a threshold for a corresponding pixel group; inputting the plurality of said generated pixel groups to a character classifier, wherein for each input pixel group, said character classifier returns an identified character and a confidence value; identifying non-overlapping input pixel groups with the highest confidence values, the identified non-overlapping pixel groups having at least two different pixel thresholds; and selecting one or more identified characters associated with said identified non-overlapping pixel groups.
2. The method according to claim 1 , wherein said non-binary image is a gray-scale or color image.
3. The method according to claim 1 , wherein said pixel intensity is a numeric value representing a specific color.
4. The method according to claim 1 , wherein said character classifier scales said generated pixel group to a specific size.
5. The method according to claim 4 , wherein said pixel group scaled to a specific size enables a matrix comparison against defined characters in said character classifier.
6. The method according to claim 1 , wherein said step of identifying non-overlapping input images with the highest confidence values segments said non-binary image into separate segments for each character.
7. The method according to claim 6 , wherein each said segment comprises a bounding rectangle consisting of a two dimensional set of adjacent pixels.
8. A computer program product for segmenting and isolating one or more characters from a non-binary image comprising one or more characters, the computer program product comprising: a non-transitory computer usable medium having computer usable code embodied therewith, the computer program product comprising: computer usable code configured for identifying a pixel intensity value of each pixel in said non-binary image; computer usable code configured for generating a plurality of pixel groups comprising contiguous pixels of equivalent or lower pixel intensity values for each pixel intensity value found in said non-binary image, each such pixel intensity value defining a threshold for a corresponding pixel group; computer usable code configured for inputting the plurality of said generated pixel groups to a character classifier, wherein for each input pixel group, said character classifier returns an identified character and a confidence value; computer usable code configured for identifying non overlapping input pixel groups with the highest confidence values, the identified non-overlapping pixel groups having at least two different pixel thresholds; and computer usable code configured for selecting one or more identified characters associated with said identified non-overlapping pixel groups.
9. The computer program product according to claim 8 , wherein said non-binary image is a gray-scale or color image.
10. The computer program product according to claim 8 , wherein said pixel intensity is a numeric value representing a specific color.
11. The computer program product according to claim 8 , wherein said character classifier scales said generated pixel group to a specific size.
12. The computer program product according to claim 8 , wherein said pixel group scaled to a specific size enables a matrix comparison against defined characters in said character classifier.
13. The computer program product according to claim 8 , wherein said step of identifying non-overlapping input pixel groups with the highest confidence values segments said non-binary image into separate segments for each character.
14. The computer program product according to claim 13 , wherein each said segment comprises a bounding rectangle consisting of a two dimensional set of adjacent pixels.
Cooperative Patent Classification codes for this invention. Click any code to explore related patents in that topic.
June 23, 2008
December 11, 2012
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.